Open-Source Automated Driving

Qualifying the future of
open-source
automated driving.

We benchmark and qualify open-source automated driving software. Every result is reproducible and aligned with ISO 26262 & 21448, so OEM and Tier-1 teams can adopt it on evidence rather than assumption.

Get in touch See a sample report
The Problem We Solve

The trust gap in open-source AD

Open-source automated driving stacks are powerful, but without transparent benchmarking and qualification, industrial adoption remains slow and risky.

Regulators

Need auditable evidence for every component and the overall system.

Internal Procurement

Need documented qualification they can sign off on.

Program Leads

Need benchmarks they trust and can defend in a review.

We close that gap. Benchmarks run on your data, and qualification documentation that survives a review.

Why Now

The evidence bar is rising

  1. 2024

    ISO/PAS 8800 published: safety and AI in road vehicles now has a dedicated standard.

  2. 2024–26

    UNECE framework expands: ALKS (R157) amended, DCAS (R171) in force and under further amendment.

  3. 2026

    UN R185 and UN GTR No. 26 adopted: the first global rules for fully driverless ADS.

    EU AI Act transparency rules start applying.

  4. 2027–28

    EU AI Act high-risk obligations phase in, Annex III from late 2027 and product-embedded AI from 2028.

What it means for type approval

Adopted by UNECE WP.29 on 24 June 2026, the framework sets the bar: an automated driving system has to be at least as safe as a competent and careful human driver. It opens two parallel routes to approval.

1958 Agreement

UN R185

A UN Regulation: international type approval with mutual recognition between contracting parties. Expected to be the route the EU builds on for driverless vehicles.

1998 Agreement

UN GTR No. 26

A global technical regulation: a common technical baseline that each contracting party transposes into national law, through type approval or self-certification.

In practice

Approval by evidence

Whether you build or buy, the questions are the same: where is the evidence, and can you reproduce it? Open-source components are no exception. They need the same qualification as anything else in your safety case.

How We Solve It

Discovery, evaluation, qualification

Automotive and robotics teams building on open-source software face one core challenge: regulators and OEM procurement teams want proof of systematic component evaluation. We make that proof possible, and repeatable.

Discover

We identify the right open-source components for your target application and compliance requirements, and map them against your existing stack.

Evaluate

We run reproducible benchmarks across real operating scenarios and safety conditions, testing performance and reliability on your data.

Qualify

We produce structured qualification documentation aligned with ISO 26262 & 21448 and relevant industry standards, ready for your internal sign-off or external audit.

Our Core Product

Autonomy.Hub

Our open-source ecosystem for evaluating automated driving software on ROS 2, from your module and your data to metrics-based evidence.

02 | Data

03 | Processing

04 | Evidence

01 | Configuration

  • Your module or full AD stack
  • Your data
  • Your deployment and ODD configuration

TestScene

Available

Autonomy.Datasets

Makes heterogeneous and fragmented data available for evaluation.

In development

Autonomy.Simulation

Runs large-scale scenario-based simulation, open or closed loop.

Autonomy.Benchmarks

Available

Autonomy.Evaluation

Applies reusable metrics across component-level and full-stack use cases.

  • Component-level example 3D object detection
  • Full-stack Criticality assessment
  • Visualization See the results in a visual format.
  • Report Detailed analysis of the benchmark results.
  • Leaderboard Compare the performance of different modules or systems.
In development

Autonomy.Bot

Orchestrates the complete benchmark chain in your CI/CD workflow.

Explore Autonomy.Hub
The Deliverable

Evidence you can hand to your auditor

Benchmark Report | Illustrative example: 3D object detection
Illustrative benchmark leaderboard ranking three lidar 3D object detection modules by NDS, with mAP, mATE, mASE, mAOE, mAVE and mAAE values and the container image used for each run.
Illustrative leaderboard view: best run per module, ranked by NDS, with each row traceable to the container image it ran on.

Benchmarks run on your data and your ODD, not on a demo set we picked because it flatters the results.

Every run is containerized and reproducible: same input, same result, re-runnable on demand.

Findings map back to requirements, so they drop straight into your safety case.

Request the sample report Book a 30-min intro call
On-Road Evidence

Developed and validated on prototype vehicles under real operating conditions, not just in simulation. When we evaluate open-source modules, the methodology behind it has already been proven in the field.

A fleet of autonomous vehicles staged on a test track at sunset.
Our Offer

Start small, scale with evidence

Three packages that build on each other. Start with a focused scouting engagement, then scale to full benchmarking and qualification.

Start here

Component Scouting

2–4 weeks | fixed price

  • Identification & mapping of suitable open-source components for your target application
  • Compliance requirements check
  • Clear adopt/avoid recommendation
Request scoping call
Core offer

Benchmark Report

6–10 weeks | fixed price

  • Reproducible, scenario-based benchmarking on your data
  • Transparent metrics, documented methodology
  • Audit-ready report
Request scoping call
Pilot slots

Qualification Package

8–14 weeks | fixed price

  • Structured qualification documentation aligned with ISO 26262 & ISO 21448
  • Ready for internal sign-off or external audit
Request scoping call

Score Board (SaaS)

Continuous CI-based evaluation, regression tracking, dashboard & API.

Coming 2027
Our Values

Open source is our foundation, not our marketing

How we work

  • Built on OpenADS using ROS 2 and Autoware compatible.
  • Every evaluation process fully documented and inspectable.
  • Scenario-based full-stack testing in simulation and on real data.

What Thinking Cars is not doing

  • Not overselling a public product that does not exist yet.
  • Not selling reports nobody can reproduce.
  • Not making claims we cannot back with evidence.
The People behind Thinking Cars

Built on deep domain expertise

The team comes out of autonomy research and software integration, and works on making open-source AD software trustworthy enough to build on.

Strategy and Technology

Dr.-Ing. Timo Woopen

CEO

PhD in collective learning for automated driving | led the UNICARagil project | responsible for strategy, offering, and customer projects.

Architecture

Raphael van Kempen

Lead Architect

PhD research in automated driving system architecture | led the autotech.agil project | responsible for the benchmark platform and project delivery.

Development

Dr.-Ing. Haohao Hu

Lead Developer

PhD in mapping for automated driving | responsible for benchmark logic, scenario design, and report quality.

Simulation

Christian Geller

Lead Simulation

PhD research in large-scale scenario-based simulation | responsible for the simulation environment and reproducible test execution.

Our Advisors

Strong network and decades of expertise in automated driving, bridging research and industry.

Prof. Dr.-Ing. Lutz Eckstein ika RWTH | VDI President

Our link into the OEM network and into industry policy.

Prof. Dr.-Ing. Christoph Stiller mrt KIT | IEEE Chair (SDA-TC)

Has shaped automated driving research and evaluation methodology for decades.

Prof. Dr.-Ing. Klaus Dietmayer mrm UULM

Sensor fusion and perception, and how to tell whether a stack really works.

Milestones

Our history and roadmap

2021 Past

Foundation

Thinking Cars GmbH is founded out of Aachen, Karlsruhe, and Ulm to turn autonomy research and systems knowledge into customer-ready delivery and clearer software decisions.

2022 Past

UNICARagil partnership

TC becomes a partner in the UNICARagil consortium, strengthening its position in the open-source automated driving space.

2022–2025

autotech.agil funding and first benchmark platform

TC joined the autotech.agil initiative and showcased a first reusable benchmark environment for structured software comparison.

2026 | Now

First open-source integration milestone

We joined the Ecosystem Mobility 4.0 initiative by the German Federal Ministry for Economic Affairs and Energy with our transparent evaluation approach. First pilot engagements are open, and we are onboarding a limited number of partners.

2027+

Open AD stack integration expands

The roadmap grows toward broader benchmarking integration, qualification logic, and delivery readiness for open-source automated driving software.

2028+

Open AD stack on the road

We form an ecosystem of well-evaluated open-source automated driving components, tested and qualified for real-world deployment.

Data & IP

Your data stays yours

Benchmarking on customer data only works with strict data discipline. Here is how we handle it.

01 Do we need an NDA first?

Yes. NDA before any data or architecture details are shared. We use your template or ours.

02 How does data hand-over work?

Via a secured data room you control, or the evaluation runs on your infrastructure and only results leave the building.

03 Is our data used for anything else?

No. Evaluation pipelines are containerized per project; your data is never reused, shared, or used to improve anything else.

04 What happens after the project?

Agreed retention period, then documented deletion. You keep the report, the methodology stays transparent.

Next Step

Be among the first to work with us.

Become a Pilot Partner

We're onboarding a limited number of pilot partners in 2026, with early-partner conditions in exchange for a joint reference. Start small: a focused component scouting, delivered in weeks, shows what our evidence looks like before you commit to a full benchmark.